ABSTRACT Alzheimer Disease (AD) is an irreversible chronic neurodegenerative disease, the most common cause of dementia among the elderly. AD is one of the biggest health problems facing our society. The cost of caring for AD patients is more than $200 billion per year in the US and is expected to increase as the population ages. Effective treatments for AD have been slow to develop due to the inability to accurately and frequently monitor AD progression in its very early stages. A strong link has been established between the appearance of excessive amounts of paired helical filament (PHF) Tau protein and the neurodegenerative process that marks dementia caused by AD. The advent of [18F] AV-1451, a tracer used in Positron Emission Tomography (PET) that selectively binds to PHF-Tau, has enabled in vivo observation. The conventional approach used to determine the rate of Tau-protein deposition in prodromal AD is to scan a subject two times (between 2-3 years apart), reconstruct images of each scan separately, and derive an annualized rate of Tau deposition for each patient on a region-by-region basis from the difference of consecutive images. Region-wise Tau accumulation is then statistically analyzed to find significant differences. The conventional approach has low sensitivity to changes in Tau deposition due to increased intensity variation in the difference image, thereby rendering the conventional approach inadequate for analysis of Tau in very early AD. We propose a joint longitudinal image reconstruction approach where the Tau deposition difference image is reconstructed directly from measurements, drastically lowering the intensity variation in the difference image. The proposed approach increases sensitivity to slight changed in Tau thereby reducing the sample size required to conduct a comparative population study, allowing more frequent scans (e.g. spaced at 6 months instead of 2+ years), more frequent diagnosis, and faster drug development.